Discover more from Cloud Native Weekly
#15 Inside S3, Unikernels and Observability cost
Explore what is happening in Cloud native ecosystem: How S3 is designed, What is Unikernels and development in this space and lastly the hot topic: Observability cost & Platform engineering.
In this newsletter, I will keep it very crisp and to the point and will talk about some very interesting development in the Cloud-native space.
Level up your game in Cloud native industry and join us along with 750+ readers.
I have also been thinking to open this newsletter for sponsorships as a few folks reached out to me to put their sponsored content here. I would like to ensure readers that sponsorships don’t influence what we share and keep it unbiased. Also, I would like to keep it less distracting as possible so only relevant sponsors will be accepted. If you want to sponsor this newsletter, please reach out to me.
Insight into operating one of the largest distributed storage - S3 — Werner recently shared some mind-blowing insights on AWS’s most popular service S3. Many of you will be surprised that S3 is now 17+ years old technology and has become a synonym for object storage. S3 processes 100M events per second and 125 billion events per day 🤯. Caution: it is a long article but highly recommended.
Datadog vs OSS (Prometheus based) — Vishal shared a study of when Datadog becomes expensive and how it is compared with OSS solutions such as Prometheus / Thanos. This comparison is based on the metrics but Datadog's offering is much more. Observability cost is one of the most trending topics currently and Datadog, the leading observability product widely used becomes exorbitantly expensive at scale. There is no doubt that It is one of the best products in the market but at what price?
Ankit from Signoz shared an article comparing Datadog with other alternatives in the market some time ago. You can read that as well to gain more insights. You can also look into various other products like Kloudfuse, Middleware, Aspecto, logz.io, OpenObserve, Honeycomb, and many more that exist and can be really confusing. If you are confused and want help selecting the right product, feel free to reach out to me. Happy to chat about this.
Opinionated Infrastructure: Platform engineering for productivity — James from Redmonk shared a video with valuable insights on how the golden path and opinionated infrastructure can help software engineering. He emphasizes why Platform teams need to simplify complexity, optimize workflows, and enhance developer productivity. I enjoyed the analogies he made while riding the bike. Redmonk has a lot more insights to keep you busy.
Unikernels making it to the news again with the upcoming launch of Kraft.cloud. The first time I heard of Unikernels was in 2018 in Container Camp AU but over the years, I did not see much traction in this technology. Last week, I got a chance to talk to Felipe, founder of Unikraft to understand the technology and development in this space. Felipe has been spending A few features which I found very useful are — fast boot time (few ms), less than 20ms cold start, serverless, hardware-level isolation, more security, and a very tiny footprint. Too good to be true, right? But that’s where the excitement is in this space. I am waiting to try it soon. They are very fast because it runs like a lightweight process instead of a container or a VM Typical use cases are — microservices, NFV, and Serverless. It is backed by an open source project so you can try it on your machine. Subscribe to the newsletter if you want to receive updates on this topic.
Kubeflow is a CNCF incubator project now — Initially developed by Google as open source MLOps open source product and used by many companies around the world. It is a highly capable MLOps platform for large-scale implementations. However, in the last few years, Kubeflow's development was a little slow when Google diverged into Vertex AI. I hope this incubation will attract more contributors and speed up the project once again. A couple of years ago, I tried Kubeflow myself and surely it is a complex one to implement and manage. There are now various other simple projects like ClearML, BentoML, ZenML, Flyte, etc in this space to explore. If you are new and want to start your MLOps journey, read this article I wrote earlier this year. Leave your comments on what platform you are using and what challenges you are facing.
AWS will be charging for public IPs. — Effective February 1, 2024, there will be a charge of $0.005 per IP per hour for all public IPv4 addresses. So, plan it accordingly if you are using a lot of public IPs.
If you find this newsletter useful, please share it with your friends and social network.
Skypilot — open source project to quickly deploy LLM and AI models in any cloud provider.
Karina - Useful when you want to build a custom platform on top of Kubernetes. Comes with a wide variety of options.
Kubernetes Volume Autoscaler — as the name suggests, you might need this if you face issues of running out of disk in your cluster or don’t want to over-allocate.
If anyone is interested in learning about DevRel, check out Daniel’s newsletter.
That’s it, folks. I would love to hear what you have been doing and if you have any interesting articles or projects to share here at Cloud Native Weekly. Do reach out to me.